[RFC 0/6] rust: pci: add config space read/write support

[Zhi Wang <zhiw@nvidia.com>]

In the NVIDIA vGPU RFC [1], the PCI configuration space access is
required in nova-core for preparing gspVFInfo when vGPU support is
enabled. This series is the following up of the discussion with Danilo
for how to introduce support of PCI configuration space access in Rust
PCI abstrations. Bascially, we are thinking of introducing another
backend for PCI configuration space access similar with Kernel::Io.

This ideas of this series are:

- Factor out a common trait IoRegion for other accessors to share the
  same compiling/runtime check like before.

- Factor the MMIO read/write macros from the define_read! and
  define_write! macros. Thus, define_{read, write}! can be used in other
  backend.

  In detail:

  * Introduce `call_mmio_read!` and `call_mmio_write!` helper macros
    to encapsulate the unsafe FFI calls.
  * Update `define_read!` and `define_write!` macros to delegate to
    the call macros.
  * Export `define_read` and `define_write` so they can be reused
    for other I/O backends (e.g. PCI config space).

- Add a helper to query configuration space size. This is mostly for
  runtime check.

- Implement the PCI configuration space access backend in PCI
  Abstractions.

  In detail:

  * `struct ConfigSpace<SIZE>` wrapping a `pdev: ARef<Device>`.
  * `IoRegion` implementation returning the device's `cfg_size`.
  * `call_config_read!` and `call_config_write!` macros bridging to
    the existing C helpers (`pci_read_config_*` /
    `pci_write_config_*`).
  * Read accessors: `read8/16/32` and `try_read8/16/32`.
  * Write accessors: `write8/16/32` and `try_write8/16/32`.

- Introduce an rust wrapper for pci_find_ext_capability(). Thus, the
  rust driver can locate the extended PCI configuration caps.

Open:

The current kernel::Io MMIO read/write doesn't return a failure, because
{read, write}{b, w, l}() are always successful. This is not true in
pci_{read, write}_config{byte, word, dword}() because a PCI device
can be disconnected from the bus. Thus a failure is returned.

- Do we still need a non-fallible version of read/write for config space?
A rust panic in accessing the PCI config space when device is
unexpectedly disconnected seems overkill.

Zhi Wang (6):
  rust: io: refactor Io<SIZE> helpers into IoRegion trait
  rust: io: factor out MMIO read/write macros
  rust: pci: add a helper to query configuration space size
  rust: pci: add config space read/write support
  rust: pci: add helper to find extended capability
  [!UPSTREAM] nova-core: test configuration routine.

 drivers/gpu/nova-core/driver.rs |   4 +
 rust/kernel/io.rs               | 132 +++++++++++++++++++++-----------
 rust/kernel/pci.rs              |  74 ++++++++++++++++++
 3 files changed, 164 insertions(+), 46 deletions(-)

-- 
2.47.3